In modern industries, data-driven fault detection and classification (FDC) systems play a crucial role in maintaining industrial security and stability. However, the security of these systems themselves is often overlooked. A significant concern is adversarial vulnerability, where data-driven machine learning models can produce incorrect predictions when exposed to maliciously altered input data. This chapter addresses this emerging security issue in data-driven FDC systems by: (1) summarizing and comparing recent and notable adversarial attack and defense methods for fault classifiers; (2) introducing new attack and defense techniques for unsupervised fault detectors; (3) developing a novel industrial adversarial security benchmark for FDC systems with a tailored robustness verification scheme under multiple norm measurement; and (4) examining and discussing the most potentially threatening attacks to FDC systems and the most effective defense techniques to counteract these threats.